www.nature.
com/scientificreports
OPEN
Received: 7 April 2017
Accepted: 13 October 2017
Published: xx xx xxxx
Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7
Clinical evaluation of corneal
changes after phacoemulsification
in diabetic and non-diabetic
cataract patients, a systematic
review and meta-analysis
Yizhen Tang1,2, Xinyi Chen1,2, Xiaobo Zhang1,2, Qiaomei Tang1,2, Siyu Liu1,2 & Ke Yao1,2
Corneal endothelium morphological abnormalities result in fluid imbalance, stromal swelling, and loss
of transparency, thus impairing visual function. Recently, growing number of studies have focused on
diabetic corneal abnormalities after cataract surgery and its comparison with non-diabetic patients,
the results remain conflicting. Thus, to evaluate the effect of phacoemulsification on the corneal
properties in diabetic and non-diabetic patients, prospective studies were comprehensively searched
through PubMed, EMBASE, and Cochrane databases updated to Jan 2017. A meta-analysis of the 13
identified studies was performed using weighted mean difference (WMD) and 95% confidence interval
(CI). For the dynamic changes between preoperative and postoperative values, significant differences
were identified between the two groups in endothelial cell density (ECD) and hexagon cells (HC%) at
1 day, 1 week, 1 month, and 3 months postoperatively, in central corneal thickness (CCT) at 1 month
postoperatively, and in coefficient variation (CV) at 1 week and 1 month postoperatively. However,
no significant differences were observed in CCT at 1 day, 1 week and 3 months postoperatively or in
CV at 1 day and 3 months postoperatively. Diabetic corneas are more vulnerable to stress and trauma,
resulting in greater morphological abnormalities and longer recovery time.
As of 2015, an estimated 415 million people had diabetes worldwide1, which is almost 1.5 times greater than in
2010 (285 million). The global prevalence of diabetes is growing much faster than earlier forecasts predicted (366
million by the year 2030)2,3. Obviously, diabetes mellitus (DM) is becoming more prevalent and threatening than
was previously thought.
It has been acknowledged that DM leads to various complications such as nephropathy, neuropathy, cardio-
vascular issues, and several ocular complications like diabetic retinopathy, diabetic cataract, diabetic keratopathy,
and diabetic optic nerve diseases4. Although the cornea may appear disease free in the diabetic, an awareness of
the marked biochemical and ultrastructural abnormalities in the diabetic enables us to prevent more overt com-
plications. The diabetic cornea suffers from endothelium cellular dysfunction and dysfunctional repair mecha-
nisms including corneal edema, delayed wound healing, and so on5,6. Over the past decades, the pathology of the
diabetic corneal endothelium dysfunction has become understood in more detail7. The state of hyperglycemia
results in an increase in aldose reductase activity, the expression of metalloproteinase (MMP), and the formation
of advanced glycation end products (AGEs). Evidence has shown that the inhibition of aldose reductase reduces
dysmorphological changes in the corneal endothelium8,9. Enhanced MMPs can damage the basement membrane
and limit cell migration, resulting in poor healing10. Moreover, the accumulation of AGEs can lead to an abnor-
mality of cell adhesion11. The cornea is likely to be more vulnerable to stress and trauma in diabetic patients than
in non-diabetics.
Phacoemulsification has become the predominant treatment procedure for cataract, the leading cause
of blindness and visual impairment worldwide12,13. Although most cataract patients achieve decent recovery,
1
Eye Center, Second Affiliated Hospital, School of Medicine, Zhejiang University, Hangzhou, Zhejiang, P.R. China.
2
Key Laboratory of Ophthalmology of Zhejiang Province, Hangzhou, P.R. China. Correspondence and requests for
materials should be addressed to K.Y. (email: xlren@zju.edu.cn)
1
www.nature.com/scientificreports/

Figure 1.  Workflow chart of the literature selection process.

unfortunately, in a complex disease environment, cataract surgery with phacoemulsification and lens implan-
tation leads to larger endothelial cell loss in diabetic corneas14–18. Furthermore, the corneal endothelium can be
adversely affected by surgery due to factors like lens nuclear sclerosis, effective phacoemulsification time (EPT),
phacoemulsification energy, and IOL implantation15,19–21. These factors coupled with the effect of DM indicate
a great risk of long-term endothelium cell dysfunction with decompensation and the development of bullous
keratopathy22. Accelerated losses of corneal endothelial cells have been reported to continue even 10 years after
surgery23.
To evaluate the corneal state, corneal thickness and endothelial cell morphology are the top two clinical con-
cerns. Central corneal thickness (CCT), as an indicator of the physiological condition of the corneal endothelium,
is generally used in diagnoses like keratoconus, Fuchs’ dystrophy, and glaucoma. Recognizing CCT is important
because it can mask an accurate reading of intraocular pressure (IOP)24, causing doctors to unnecessarily treat
for a condition that may not exist. Endothelial morphological changes in corneas, including endothelial density
(ECD), coefficient of variation (CV), and percentage of hexagonal cells (HC%), can alter the cornea’s ability to
function. Abnormal corneal endothelial cell morphology coupled with increased CCT25 is another marker of
endothelial cell dysfunction, which results in fluid imbalance, stromal swelling, and loss of transparency, thus
impairing visual function.
Recently, a growing number of studies have focused on the importance of diabetic corneal abnormalities,
which were commonly found in patients after cataract surgery due to factors including diabetic state and surgical
procedures26. However, there is still conflict concerning the differences in corneal properties between diabetic and
non-diabetic patients after phacoemulsification. According to our knowledge, there has been no comprehensive
review or meta-analysis regarding corneal changes after phacoemulsification for diabetic and non-diabetic groups
so far. Under the circumstances, this article is set to evaluate the effect of phacoemulsification on ECD, HC%, CV,
and CCT in diabetic patients and non-diabetic patients.

Results
Literature selection.  The workflow chart in Fig. 1 shows the literature selection process. After duplication
removal, a total of 361 studies were retrieved from the databases. 330 studies were excluded by scanning titles and
abstracts. Furthermore, 18 studies were excluded after full-text reading: four on extracapsular cataract extraction
(ECCE), two on manual small incision cataract surgery (MSICS) which is different from phacoemulsification
from either the incision size or the surgical method, five retrospective studies, five with unavailable data (e.g. no
postoperative data, no cohort, or an unmatched comparison group), one that was a poster, and one that was a
review. Finally, 13 studies17,19,27–37 meeting all of the predefined criteria were identified. The characteristics of the
included studies are shown in Table 1. The details of N OS scale are shown in Table S6.

Meta-analysis of the outcomes.  A total of 13 prospective studies including 1923 eyes (941 in the
non-diabetic group and 982 in the diabetic group) were identified. The surgical parameters (e.g. phaco-time and
phaco-energy) and lens nucleus hardness were reported with no significant differences between the DM and
non-DM groups, as shown in Tables S1–S4.

Endothelial cell density.  There were thirteen studies reporting the outcome of ECD. The analysis was
made at 1 day, 1 week, 1 month, and 3 months postoperatively. It was found that diabetic patients have a sig-
nificantly lower ECD at preoperative and all postoperative time points than the non-diabetic group (baseline:
WMD = −98.60, 95% CI: −181.39 to −15.82, P = 0.02; 1 day postoperative: WMD = −129.29, 95% CI: −149.47

Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7 2

www.nature.com/scientificreports/

No. of eyes Gender

(M/F) Age (year/SD, range) Quality
Follow up control
Study ID Country/province Study design duration Normal DM Normal DM Diabetes condition (NOS)
Prospective
Li19
China Hainan 3 months 106/84 101/85 65.1 (12.3) 64.6 (12.5) NA 7
controlled study
Prospective Duration (11.54 years) HbA1c%
Misra27 New Zealand 6 months 23 28 74.4 (7.4) 71.2 (7.6) 8
cohort study (DM:nDM 7.5:5.7, P < 0.001)
Prospective Blood glucose < 8.3 mmol/L,
Zhu28 China Beijing 1 month 16/17 15/20 60–80 60–80 8
controlled study diabetic course > 5a
Prospective Diabetic course > 10a, Diabetic
Yan29 China Guizhou 3 months 30/32 51/46 69.6 (55–85) 65.6 (52–82) 8
controlled study course < 10a
Prospective Blood glucose < 8.3 mmol/L,
Su30 China Liaoning 1 month 31 39 50–80 50–80 8
controlled study diabetic course < 15a
Prospective
Liu31 China Hebei 1 week 16/14 17/13 64.7 (3.6) 69.6 (5.4) NA 8
controlled study
Prospective
Zhao32 China Shandong 3 months 24/26 26/24 64.8 (55–83) 63.2 (52–80) NA 8
controlled study
Prospective
Wang33 China inner Mongolia 6 months 36/24 24/16 69.2 (8.2) 65.3 (11) NA 7
controlled study
Prospective Blood glucose < 6. 6–10 mmol/L
Yang34 China Guangzhou 3 months 33/32 34/41 68.1 (50–84) 67.8 (51–83) 8
controlled study Duration (1–17 years)
Prospective
Hugod35 Denmark 3 months 30 30 75.4 (9.3) 75.6 (8.6) HbA1c%(7.08 (1.43)) 8
controlled study
Prospective
Wu36 China Shaanxi 3 months 31 28 50–80 50–80 Diabetic course < 12a 7
controlled study
Prospective
Wu37 China Hainan 3 months 33/31 26/24 65.6 (52–82) 66.8 (50–84) NA 8
controlled study
Prospective
Morikubo17 Japan 1 month 93 93 68.8 (8.9) 68.6 (8.8) NA 8
controlled study

Table 1.  Characteristics of included studies. NA = not available, DM = diabetes mellitus, NOS = The
Newcastle–Ottawa scale.

to −109.10, P < 0.001; 1 week postoperative: WMD = −192.17, 95% CI: −267.04 to −117.29, P < 0.001; 1
month postoperative: WMD = −205.53, 95% CI: −258.30 to −152.76, P < 0.001; 3 months postoperative:
WMD = −229.83, 95% CI: −283.54 to −176.12, P < 0.001; Fig. 2). Furthermore, the percentage of the loss of
ECD (ECL%, difference between preoperative and postoperative), which was calculated from equations (1–5),
was also evaluated to see the effect caused by phacoemulsification. There are significant differences in ECL% at
all postoperative times for the DM group compared to the non-DM group (1 day postoperative: WMD = 3.40,
95% CI: 1.82 to 4.97, P < 0.001; 1 week postoperative: WMD = 3.45, 95% CI: 2.64 to 4.25, P < 0.001; 1 month
postoperative: WMD = 3.80, 95% CI: 1.84 to 5.75, P < 0.001; 3 months postoperative: WMD = 4.85, 95% CI: 1.60
to 8.10, P = 0.003; Fig. 3).

Coefficient of variation.  There were eight studies reporting the outcome of coefficient of variation. The
analysis was made at 1 day, 1 week, 1 month, and 3 months postoperatively. It was observed that diabetic patients
had a significantly higher CV at preoperative and all postoperative time points than the non-DM group (baseline:
WMD = 2.62, 95% CI: 1.41 to 3.83, P < 0.001; 1 week postoperative: WMD = 5.09, 95% CI: 2.68 to 7.51, P < 0.001;
1 month postoperative: WMD = 6.71, 95% CI: 3.60 to 9.81, P < 0.001; 3 months postoperative: WMD = 6.65,
95% CI: 3.14 to 10.15, P < 0.001), except at 1 day postoperatively (WMD = 4.75, 95% CI: −1.51 to 11.00, P = 0.14,
Fig. 4). The increase of CV (dCV, difference between preoperative and postoperative) appears significantly larger
in the DM group compared to the non-DM group at 1 week postoperatively (WMD = 2.02, 95% CI: 0.66 to 339,
P = 0.004) and at 1 month postoperatively (WMD = 3.68, 95% CI: 0.47 to 6.88, P = 0.02), while no significant
differences were found at 1 day postoperatively (WMD = 3.27, 95% CI: −3.02 to 9.56, P = 0.31) and 3 months
postoperatively (WMD = 3.24, 95% CI: −1.06 to 7.53, P = 0.14, Fig. 5).

Hexagonal cell percentage.  There were 11 studies reporting the outcome of hexagonal cells. The analysis
was made at 1 day, 1 week, 1 month, and 3 months postoperatively. It was found that diabetic patients have a sig-
nificantly smaller HC% at preoperative and all postoperative time points (all: P < 0.001, Fig. 6) and a significantly
larger HC% loss (difference of preoperative and postoperative) at all postoperative times (all: P < 0.001, Fig. 7)
compared to the non-DM group.

Central corneal thickness.  There were four studies reporting the outcome of CCT. The analysis was made
at 1 day, 1 week, 1 month, and 3 months postoperatively. No significant difference was observed in CCT between
diabetic patients and non-diabetic patients. (baseline: WMD = 2.86, 95% CI: −3.65 to 9.37, P = 0.39, 1 day post-
operative: WMD = 13.37, 95% CI: −5.24 to 31.99, P = 0.16). After phacoemulsification, the CCT of the DM
group was significantly higher than that of the non-DM group at all postoperative time points (1 week postoper-
atively: WMD = 17.96, 95% CI: 5.24 to 30.68, P = 0.006; 1 month postoperatively: WMD = 22.59, 95% CI: 10.23
to 34.94, P < 0.001; 3 months postoperatively: WMD = 12.92, 95% CI: 9.22 to 16.63, P < 0.001; Fig. 8). Moreover,

Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7 3

www.nature.com/scientificreports/

Figure 2.  Forest plot comparison of the corneal endothelial cell density (ECD) after phacoemulsification in
diabetic and non-diabetic patients.

the percent increase of CCT (dCCT%, difference between preoperative and postoperative percentages) showed
no significant difference between the DM group and the non-DM group at 1 day postoperatively (WMD = 2.26,
95% CI −1.82 to 6.43, P = 0.28), 1 week postoperatively (WMD = 2.81, 95% CI: −0.36 to 5.98, P = 0.08) and 3
months postoperatively (WMD = 1.56, 95% CI: −0.57 to 3.70, P = 0.15), but a significantly larger dCCT% was
found in diabetic patients at 1 month postoperatively compared to the non-diabetic ones (WMD = 3.86, 95% CI:
1.28 to 6.45, P = 0.003, Fig. 9).

Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7 4

www.nature.com/scientificreports/

Figure 3.  Forest plot comparison of the corneal endothelial cell loss in percentage (ECL%) after
phacoemulsification in diabetic and non-diabetic patients.

Sensitivity analysis and publication bias.  Since some of the results show heterogeneity (I 2 > 50), a
random-effect meta-regression model was chosen to analyze this condition. No publication bias was found
through Begg’s and Egger’s test as shown in Table S5. Outcomes included in only 3 studies are too small to do a
sensitivity analysis so as to exclude trials at a high risk of bias. Thus, one-study-removed analyses were conducted
for all remaining outcomes. The sensitivity analysis revealed that there are two pooled outcomes lacking stability:
the outcome of ECD at the baseline and the outcome of dCV at 1 month postoperatively. Yan and Chen (2014)29,
whose patients had a long duration of DM, is the source of the statistical heterogeneity in the meta-analysis of the
preoperative ECD result as shown in Table 2. When this outlier study is removed, the outcome is stable, indicating

Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7 5

www.nature.com/scientificreports/

Figure 4.  Forest plot comparison of the coefficient of variation (CV) of corneal endothelial cells after
phacoemulsification in diabetic and non-diabetic patients.

that the outcome is still rational and reliable. The heterogeneity of the second sensitive outcome may be due to
several design differences among the studies that affect the response degree and recovery duration, such as dura-
tion of diabetes and blood glucose control. No significant publication bias was demonstrated in the funnel plot.

Discussion
The results of the present meta-analysis provided robust evidence that the effect of phacoemulsification on cor-
neal changes in diabetics is greater than for non-diabetics. Significant differences have been observed between
the diabetic and non-diabetic groups in terms of ECD, HC%, CV, and CCT preoperatively and 1 day, 1 week,

Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7 6

www.nature.com/scientificreports/

Figure 5.  Forest plot comparison of the change of coefficient of variation (dCV) of corneal endothelial cells
after phacoemulsification in diabetic and non-diabetic patients.

1 month, and 3 months postoperatively, except CV at 1 day postoperatively and CCT preoperatively. For the
changes between the preoperative and postoperative state, significant differences were identified in ECL% and
HC% loss at 1 day, 1 week, 1 month, and 3 months postoperatively, in dCCT% at 1 month postoperatively, and in
dCV at 1 week and 1 month postoperatively between the two groups. Nevertheless, no significant differences were
observed in dCV and dCCT% at 1 week and 3 months postoperatively.

Corneal endothelial cell morphology.  It was once considered controversial that diabetes could affect
corneal endothelium morphology preoperatively. Inoue et al.38 investigated 1394 patients before cataract surgery
and their multiple regression analysis revealed that age instead of DM was the only variable relevant to ECD,
CV, and HC%. However, Lee et al.39 reported that corneal endothelium morphology was significantly different
between DM and non-DM patients, and CV is significantly correlated with diabetes duration. Taking multiple
studies into account, the results of this review offer the judgment that DM patients have lower ECD and HC%,
but higher CV, than non-DM patients (CV: P < 0.001; ECD: P = 0.02; HC%: P < 0.001; Figs 2, 4 and 6) before
phacoemulsification.

Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7 7

www.nature.com/scientificreports/

Figure 6.  Forest plot comparison of the percentage of hexagonal corneal endothelial cells (HC%) after
phacoemulsification in diabetic and non-diabetic patients.

The fragility of the corneal endothelium in the eyes of diabetic patients might be explained by several mech-
anisms. With an enhanced polyol pathway and the accumulated sugar alcohol in cells converted by excessive
glucose, the osmotic pressure goes up, causing the fragility of diabetic corneal endothelial cells. Morphological
abnormalities in the corneal endothelium have been reported8,40 to improve after administration of an aldose
reductase inhibitor in the polyol pathway, which supports its involvement in the corneal endothelial abnormali-
ties of patients with DM. Furthermore, the enhanced accumulation of AGEs in diabetic corneas provides strong
evidence22 that nuclear oxidative DNA damage caused by the accumulation of AGEs is responsible for the apop-
totic damage of corneal endothelial cells in diabetic patients, which also results in decreased ECD. Diabetes also

Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7 8

www.nature.com/scientificreports/

Figure 7.  Forest plot comparison of the loss of percentage of hexagonal corneal endothelial cells (dHC%) after
phacoemulsification in diabetic and non-diabetic patients.

reduces the activity of Na+/K+-ATPase of the corneal endothelium41, which plays a key role in the maintenance of
its structure. This causes morphological and functional changes, including increased CV and decreased HC% in
diabetic corneas. Since a regular hexagonal pattern of the corneal endothelium provides the most stable covering
plane, deviation from this pattern leads to a less stable monolayer. In the diabetic endothelium, there is a greater
surface tension on the monolayer caused by the loss of the regular hexagonal pattern and the increasingly irregu-
lar shapes of the corneal endothelium, which makes the corneas of diabetic patients more fragile26,42. According
to quantitative morphometric analysis, the corneas of patients with DM may be at risk in intraocular surgical
procedures.

Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7 9

www.nature.com/scientificreports/

Figure 8.  Forest plot comparison of the central corneal thickness (CCT) after phacoemulsification in diabetic
and non-diabetic patients.

Endothelial cell density.  Clinical observations have indicated that the corneal endothelium is capable of
compensation to prevent complicated diseases of the cornea, such as bullose keratopathy22, unless the cell density
reaches a very low threshold of 400–500 cells/mm2, at which point the cornea cannot maintain its normal physi-
ological function43. Typically, it is widely accepted that 1000 cells/mm2 is the minimum preoperative value to pre-
vent corneal decompensation after surgery. The diabetic cornea, which is more fragile and vulnerable to trauma,
possesses a weaker compensatory capacity. The study by Furuse et al.44 reportedfwhich point the cornea cannot
that there is no significant difference in the ECL% and CV values beween the two groups postoperatively.
However, a recent study by Dhasmana et al.45 showed a severe increase in ECL% in the DM group compared to
the control after cataract surgery. The results of our meta-analysis showed that DM patients have a significantly
greater ECL% than non-DM patients from the first day to 3 months postoperatively (P < 0.01, Fig. 3), confirming
that diabetic patients are more susceptible to corneal endothelial damage after phacoemulsification.

Coefficient of variation and hexagonal cell percentage.  It’s common practice that ECD is used to
evaluate the state of corneas after phacoemulsification, but it cannot reflect the dynamic of the healing process for
trauma. The change in morphology has a closer relationship with the dynamic of the corneal recovery process.
The loss of endothelial cells, as an immediate response to surgery, leads to some defeats. Unlike the corneal epi-
thelium, the cells of the endothelium do not regenerate. Instead, the remaining cells enlarge and stretch to cover
the posterior corneal surface in order to fill the space. Ideally, the earliest phenomena should be an increase in

Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7 10

www.nature.com/scientificreports/

Figure 9.  Forest plot comparison of the increased central corneal thickness percentage (dCCT%) after
phacoemulsification in diabetic and non-diabetic patients.

cell size coupled with an enlargement in CV and a decrease in HC%. After a period of rearrangement, the defects
would diminish, and CV and HC% would return to preoperative values as well46. Gradually, the cells return to
stability to maintain the physiological function of corneas. Although many studies28–30,32,34–37,45,47 have mentioned
that a significant difference in CV and HC% were found between DM and non-DM groups or between postoper-
ative and preoperative periods, quite a few studies discuss the comparison of the dynamic change of CV and HC%
between the two groups. Our meta-results showed that dCV follows a pattern of maximum increase between
1 day and 1 week postoperatively and then slowly reduces for at least 3 months. On the first day after phacoemul-
sification, both groups have an enlarged inhomogeneity in cell size, which gives two large dCV values without
a significant difference (P = 0.31). The ascending process lasts for 1 day17,34 to 1 month32 before cell shape starts
to compensate to be uniform. Significant differences between the two groups started at 1 week postoperatively
(P = 0.004), peaked at 1 month postoperatively (P = 0.02, Fig. 7), and then vanished at 3 months postoperatively
with the recovery of the diabetic patient (P = 0.14, Fig. 5). However, a sensitivity analysis showed an unstable
outcome at 1 month postoperatively. It can be inferred that this is because the measurements were done at some
critical point when the significant difference started to appear or disappear. What’s more, different durations of
DM29 and blood glucose control34,48 might affect the result.
This kind of recovery was not uniform for all corneal morphological properties. For HC% loss, there was
always a significant difference between the DM and non-DM groups postoperatively (P < 0.001, Fig. 7). Once
the endothelial cells lose their hexagonal structure, the stretching and rearrangement process will not easily give
a second chance for the cells to stabilize into hexagons again. The time of the HC% recovery in diabetic corneas
should be longer than 3 months.

Central corneal thickness.  For the preoperative state, Kotecha et al.49 found no significant difference in
CCT between the DM and non-DM groups. However, Lee et al.18,39 reported that CCT, which is strongly corre-
lated with DM duration, is slightly greater in diabetic patients than in non-diabetic patients. The analysis revealed
a slightly greater CCT in patients with DM, but no significant difference preoperatively (P = 0.39, Fig. 8).
The effect of DM on CCT is still ambiguous. There are several possible explanations, such as the inhibition
of endothelial pumping, growing stromal swelling pressure, and increasing endothelial permeability caused by

Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7 11

www.nature.com/scientificreports/

Study P 95% CI I2
Endothelial cell density preoperatively
Hugod35 P = 0.01 [−193.35, −21.02] 93%
Li19 P = 0.04 [−197.90, −4.74] 94%
Liu31 P = 0.02 [−192.48, −18.93] 93%
Misra27 P = 0.03 [−182.39, −11.18] 94%
Morikubo17 P = 0.02 [−195.75, −19.36] 93%
Su30 P = 0.03  [−189.02, −12.17] 94%
Wang33 P = 0.05  [−189.35, 1.16] 93%
Wu37 P = 0.03  [−202.90, −12.37] 91%
Wu36 P = 0.03  [−184.39, −10.50] 93%
Yan29 P = 0.01  [−100.10, −12.55] 67%
Yang34 P = 0.02  [−193.79, −18.58] 93%
Zhao32 P = 0.02  [−190.32, −13.07] 94%
Zhu28 P = 0.03 [−184.61, −9.14] 93%
Changes of coefficient of variation at 1 month postoperatively
Li19 P = 0.09 [−0.50, 7.73] 97%
Morikubo17 P = 0.009 [1.09, 7.64] 97%
Wu36 P = 0.003 [1.59, 7.93] 97%
Yan29 P = 0.01 [0.99, 7.58] 96%
Yang34 P = 0.09 [−0.45, 6.33] 94%
Zhan32 P = 0.16 [−0.97, 5.94] 97%
Zhu28 P = 0.08 [−0.34, 6.90] 97%

Table 2.  One-study-removed analysis for the outcomes of ECD preoperatively and dCV at 1 month
postoperatively.

diabetic metabolism50–53. Briefly, the normal corneal endothelium plays a key role in keeping the cornea moist
and transparent as well as maintaining integrity to prevent stromal swelling. Tight apical junctions on the
endothelial cells function as physical barriers. The movement of water outward from the corneal stroma into the
anterior chamber is increased due to ion pumps in the endothelial cells. Thus, corneal edema can be caused by a
breakdown of either the anatomical barrier or the pump function of the corneal endothelial cells, representing
an increase in CCT. This effect depends on the pathology insults of DM18 and the severity of the physical trauma.
After phacoemulsification, the increase in CCT was maximum at 1 day and 1 week postoperatively and then
gradually decreased for at least 3 months17,26,32,33,45,47. Altintas et al.26 demonstrated that corneal thicknesses were
greater in both diabetic and non-diabetic patients 1 week postoperatively than in later follow up, while there
were no differences in corneal thickness according to phaco-time or diabetic status. Nevertheless, most stud-
ies17,32,33,45,47 mentioned a delayed recovery of postoperative corneal edema in diabetics compared to normal
controls. This analysis observed significant differences between the two groups at all postoperative times from
1 day to 3 months. For the postoperative changes, the dCCT% followed the same trend as dCV. Severe postop-
erative responses and long recovery times in the diabetic group are shown by the analysis in Fig. 3. There was a
significant difference found at 1 month postoperatively. It can be inferred that this is because, at an early time
after phacoemulsification, both DM and non-DM patients have severe responses and sharp increases due to the
breakdown of corneal endothelial function caused by the surgical procedure, thus making the difference between
the two groups too small to be distinguished (P = 0.28, P = 0.08, Fig. 9). Furthermore, studies have proven that
hyperglycemia enhance the expression of MMPs54, the production and activity of which is likely to damage the
basement membrane, including type IV collagen, and limit epithelial cell migration, resulting in poor epithelial
healing55. Gradually, the corneas of non-diabetic patients start to heal more quickly, showing a smaller CCT com-
pared to the DM group, especially at 1 month postoperatively (P = 0.003, Fig. 9). Finally, differences due to the
effect of surgery vanish 3 months postoperatively (P = 0.15, Fig. 9), which means DM patients can take almost 3
months to recover.

Visual acuity.  Besides all of the changes in the cornea, visual rehabilitation is still the top concern for patients
undergoing phacoemulsification. Best corrected visual acuity (BCVA) is one of the best parameters for evaluating
the quality and efficiency of a surgical technique. Although CCT values were significantly different between the
two groups after phacoemulsification, there was no difference in visual acuity in the long-term comparison as
reported27,33,45. However, the BCVA of the non-diabetic group was better at 1 week postoperatively33, indicating
that the diabetic achieve worse vision recovery, which is consistent with the CCT results. Eventually, patients
in both groups had better postoperative visual acuity at the end of the follow-up period, which indicates that
phacoemulsification should be considered as a safe procedure for cataract extraction in the diabetic.
As a consequence, greater efforts and concentration should be made by the surgeon to minimize surgical
trauma, especially for the diabetic. To achieve that, phaco-power near the cornea should be avoided. A vis-
coelastic agent could be generously used to cushion the endothelium as well. What’s more, close postoperative

Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7 12

www.nature.com/scientificreports/

observation and intervention is suggested in patients with transient corneal edema and decompensation, as it
can be a predicted factor for the development of pseudophakic cystoid macular edema56. Since the duration of
DM and blood glucose have proven to be associated with the severity of corneal damage caused by phacoemul-
sification29,48, diabetic patients are recommended to choose the proper timing when good glycemic and HbA1c
control is achieved for cataract surgery35,48, thereby preventing further complications and minimizing visual
loss. Femtosecond laser-assisted cataract surgery is currently reported to be safer and more effective in reducing
endothelial cell loss and postoperative central corneal thickening and achieving better visual and refractive out-
comes compared to conventional phacoemulsification surgery. Therefore, it might be a potentially better choice
for the diabetic. Further studies are needed to validate this hypothesis.
To the best of our knowledge, this is the first meta-analysis and review of corneal changes after phacoemul-
sification in diabetic and non-diabetic patients. Not only the postoperative state but also the changes between
preoperative and postoperative states, which have an advantage over the former, are evaluated. In addition, the
dynamic healing process and the changes of these parameters are carefully demonstrated in this study with at least
three studies in each analysis to give a rational analysis result. We offer a systematic evaluation as well as possible
mechanisms and treatment to the clinic.
Inevitably, this meta-analysis has several limitations. First, the limitations came from the clinical trial itself.
It was not possible to have a randomized control trial (RCT) as diabetic patients are aware of their condition and
diabetic group naturally exists. Cohort studies, which are not as reliable as RCTs, were therefore included in this
meta-analysis. Secondly, the duration of diabetes for the diabetic patients, the surgical conditions, the surgeons,
and the data collection techniques all work together to make some of the outcomes not uniform. Thirdly, most
of the patients in the included studies are from Asia, which might be a potential source of deviations as corneal
biomechanics may vary among races57,58.

Method
Search strategy.  PubMed, EMBASE, and the Cochrane Controlled Trials Register were searched for rel-
evant literature dated up to Jan 2017. The following terms were used to search for prospective studies in the
selected databases: ((diabete OR diabetes) AND cataract surgery) AND corneal; ((‘cataract extraction’/exp OR
‘cataract extraction’) AND ‘diabetes mellitus’/exp OR ‘diabetes mellitus’) AND corneal; “diabetes” AND “cataract
surgery” AND “corneal”. The bibliographies of the relevant review and original research articles were also scanned
for potential trials that may have been missed in the primary searches.

Study selection.  The inclusion criteria for our selection process were set as follows: 1) Prospective con-
trolled study, 2) the study included DM patients and normal patients who underwent phacoemulsification and
IOL implantation, 3) the study reported at least one basic dataset of corneal properties, such as ECD, CV, HC%,
and CCT, 4) the patients in the trials were absent of additional underlining diseases or eye disorders other than
diabetes and cataract. Patients with other complications that could affect corneal state (e.g. severe liver or kidney
dysfunction, glaucoma, iritis, or eye injury) were excluded.

Screening process.  Two reviewers, working independently from each other, first conducted preliminary
reviews of the titles and abstracts; then, the full articles were analyzed to select the studies that met our predefined
criteria. Disagreements between the two reviewers were resolved through careful discussion, involving a third or
fourth reviewer when necessary, until a consensus was reached.

Quality assessment.  The Newcastle–Ottawa scale (NOS) was used for quality assessment. The NOS con-
tains eight items (nine scores in total) which fit into three categories: selection (four scores), comparability (two
scores), and exposure of a case-control study or outcome of a cohort study (three scores). A score ≥6 indicates
good quality.

Data extraction process.  The patient data was extracted from the selected studies via a standard form: first
author, country (province), year of publication, age of patient, sex of patient, follow up duration, quality control,
and preoperative diabetes condition. The second reviewer double-checked all data.
The measurement of corneal properties included corneal endothelial cell density (ECD), corneal endothe-
lial hexagon percentage (HC%), corneal endothelial coefficient of variation (CV), and central corneal thickness
(CCT). Most of the studies only reported the absolute values of the outcomes at a preoperative baseline and
postoperative time points. The mean and standard deviation outcomes of the corneal changes were calculated as
follows:
∆mean = meanpost − meanpre (1)

2 2
SD∆mean = SDpre + SDpost − 2ρSDpre SDpost (2)
where ρ is the covariance coefficient, pre and post are short for preoperative and postoperative state. Generally, ρ
was treated as ~0.5. The mean change percentages were calculated according to:
∆mean SD
∆mean% = , SD∆mean % = ∆mean
meanpre meanpre (3)
For situations when the selected study included multiple groups, a group-combining method was used in59

this meta-analysis to create a single pair-wise comparison. Considering a two-group combining process in which

Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7 13

www.nature.com/scientificreports/

group 1 has a sample size of N1, a mean outcome of M1, and a standard derivation of SD1, and group 2 has similar
(N2, M2, and SD2), the combined group 1 + 2 would be calculated as:
N1M1 + N2M2
N1 +2 = N1 + N2; M1 +2 =
N1 + N2 (4)

N1N2
(N1 − 1)SD12 + (N2 − 1)SD22 + N1 + N2
(M1 − M2)2
SD1 +2 =
N1 + N2 − 1 (5)
If there were more than two groups to combine, the strategy was to repeat this method sequentially (i.e.
combine group 1 and group 2 to create group 1 + 2, and then combine group 1 + 2 and group 3 to create group
1 + 2 + 3, and so on).

Statistical analysis.  The statistical analysis was performed using Review Manager 5.3. The weighted
mean difference (WMD) and 95% confidence interval (CI) were calculated from selected outcomes. P < 0.05
was considered statistically significant. Statistical heterogeneity was tested using the chi-squared and I2 tests. A
random-effect meta-regression model was used when significant heterogeneity (I2 >  50%) or clinical divergence
were found. Otherwise, a fixed-effect meta-regression model was chosen. Publication bias was measured in Begg’s
and Egger’s test using Stata 14. To evaluate the stability and reliability of our pooled outcomes, a sensitivity analy-
sis was performed using a one-study-removed analysis to assess whether the results were affected by the excessive
weight of a single study.

References
1. International Diabetes Federation (IDF). IDF Diabetes Atlas 7th Edition. http://www.diabetesatlas.org/ (2015).
2. Wild, S., Roglic, G., Green, A., Sicree, R. & King, H. Global Prevalence of Diabetes: estimates for the year 2000 and projections for
2030. Diabetes Care 27, 1047 (2004).
3. Shaw, J. E., Sicree, R. A. & Zimmet, P. Z. Global estimates of the prevalence of diabetes for 2010 and 2030. Diabetes Res Clin Pract 87,
4–14, https://doi.org/10.1016/j.diabres.2009.10.007 (2010).
4. Alberti, K. G. & Zimmet, P. Z. Definition, diagnosis and classification of diabetes mellitus and its complications. Part 1: diagnosis
and classification of diabetes mellitus provisional report of a WHO consultation. Diabetic medicine: a journal of the British Diabetic
Association 15, 539–553, https://doi.org/10.1002/(sici)1096-9136(199807)15:7<39::aid-dia668<3.0.co;2-s (1998).
5. Srinivas, S. P. Dynamic Regulation of Barrier Integrity of the Corneal Endothelium. Optometry and vision science: official publication
of the American Academy of Optometry 87, E239–E254, https://doi.org/10.1097/OPX.0b013e3181d39464 (2010).
6. Lutty, G. A. Effects of diabetes on the eye. Investigative ophthalmology & visual science 54, Orsf81–87, https://doi.org/10.1167/
iovs.13-12979 (2013).
7. Browning, D. J. Diabetic Retinopathy Evidence-Based Management. (Springer New York, 2010).
8. Matsuda, M. et al. The effects of aldose reductase inhibitor on the corneal endothelial morphology in diabetic rats. Current eye
research 6, 391–397 (1987).
9. Datiles, M. B., Kador, P. F., Kashima, K., Kinoshita, J. H. & Sinha, A. The effects of sorbinil, an aldose reductase inhibitor, on the
corneal endothelium in galactosemic dogs. Investigative ophthalmology & visual science 31, 2201–2204 (1990).
10. Ramamurthi, S., Rahman, M. Q., Dutton, G. N. & Ramaesh, K. Pathogenesis, clinical features and management of recurrent corneal
erosions. Eye (London, England) 20, 635–644, https://doi.org/10.1038/sj.eye.6702005 (2006).
11. Ljubimov, A. V. et al. Human corneal epithelial basement membrane and integrin alterations in diabetes and diabetic retinopathy.
The journal of histochemistry and cytochemistry: official journal of the Histochemistry Society 46, 1033–1041 (1998).
12. Khairallah, M. et al. Number of People Blind or Visually Impaired by Cataract Worldwide and in World Regions, 1990 to
2010Worldwide Prevalence of Cataract, 1990–2010. Investigative ophthalmology & visual science 56, 6762–6769, https://doi.
org/10.1167/iovs.15-17201 (2015).
13. Lee, C. M. & Afshari, N. A. The global state of cataract blindness. Current opinion in ophthalmology 28, 98–103, https://doi.
org/10.1097/icu.0000000000000340 (2017).
14. Diaz-Valle, D., Benitez del Castillo Sanchez, J. M., Castillo, A., Sayagues, O. & Moriche, M. Endothelial damage with cataract surgery
techniques. Journal of cataract and refractive surgery 24, 951–955 (1998).
15. Walkow, T., Anders, N. & Klebe, S. Endothelial cell loss after phacoemulsification: relation to preoperative and intraoperative
parameters. Journal of cataract and refractive surgery 26, 727–732 (2000).
16. Dick, H. B., Kohnen, T., Jacobi, F. K. & Jacobi, K. W. Long-term endothelial cell loss following phacoemulsification through a
temporal clear corneal incision. Journal of cataract and refractive surgery 22, 63–71 (1996).
17. Morikubo, S., Takamura, Y., Kubo, E., Tsuzuki, S. & Akagi, Y. Corneal changes after small-incision cataract surgery in patients with
diabetes mellitus. Archives of ophthalmology (Chicago, Ill.: 1960) 122, 966–969, https://doi.org/10.1001/archopht.122.7.966 (2004).
18. Lee, J. S., Lee, J. E., Choi, H. Y., Oum, B. S. & Cho, B. M. Corneal endothelial cell change after phacoemulsification relative to the
severity of diabetic retinopathy. Journal of cataract and refractive surgery 31, 742–749, https://doi.org/10.1016/j.jcrs.2004.09.035
(2005).
19. Li, M. H., Fu, X. L. & Yang, W. F. Effect and risk factors for corneal endothelial cells after phacoemulsification in diabetic cataract
patients. International Eye Science 16, 1048–1051 (2016).
20. Beltrame, G., Salvetat, M. L., Driussi, G. & Chizzolini, M. Effect of incision size and site on corneal endothelial changes in cataract
surgery. Journal of cataract and refractive surgery 28, 118–125 (2002).
21. Chen, X., Chen, K., He, J. & Yao, K. Comparing the Curative Effects between Femtosecond Laser-Assisted Cataract Surgery and
Conventional Phacoemulsification Surgery: A Meta-Analysis. PLoS ONE 11, e0152088, https://doi.org/10.1371/journal.
pone.0152088 (2016).
22. Bikbova, G., Oshitari, T., Tawada, A. & Yamamoto, S. Corneal changes in diabetes mellitus. Current diabetes reviews 8, 294–302
(2012).
23. Bourne, R. R. et al. Effect of cataract surgery on the corneal endothelium: modern phacoemulsification compared with extracapsular
cataract surgery. Ophthalmology 111, 679–685, https://doi.org/10.1016/j.ophtha.2003.07.015 (2004).
24. Doughty, M. J. & Zaman, M. L. Human corneal thickness and its impact on intraocular pressure measures: a review and meta-
analysis approach. Survey of ophthalmology 44, 367–408 (2000).

Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7 14

www.nature.com/scientificreports/

25. Lundberg, B., Jonsson, M. & Behndig, A. Postoperative corneal swelling correlates strongly to corneal endothelial cell loss after
phacoemulsification cataract surgery. American journal of ophthalmology 139, 1035–1041, https://doi.org/10.1016/j.ajo.2004.12.080
(2005).
26. Altintas, A. G., Yilmaz, E., Anayol, M. A. & Can, I. Comparison of corneal edema caused by cataract surgery with different phaco
times in diabetic and non-diabetic patients. Annals of ophthalmology (Skokie, Ill.) 38, 61–65 (2006).
27. Misra, S. L., Goh, Y. W., Patel, D. V., Riley, A. F. & McGhee, C. N. J. Corneal microstructural changes in nerve fiber, endothelial and
epithelial density after cataract surgery in patients with diabetes mellitus. Cornea 34, 177–181 (2015).
28. Zhu, N., Zhang, Z. C. & Hao, X. L. Influence of phacoemulsification on corneal endothelial cell of cataract patients with diabetes or
hypertension. International Eye Science 14, 480–483 (2014).
29. Yan, A. M. & Chen, F. H. Phacoemulsification on corneal endothelium cells in diabetic patients with different disease duration.
International Eye Science 14, 1786–1789 (2014).
30. Su, C. & Liu, D. Clinical observation of changes of corneal endothelial cells before and after the cataract ultrasonic emulsification for
diabetics. International Eye Science 14, 273–275 (2014).
31. Liu, J. et al. Study on corneal endothelial cells after phacoemulsification. International Eye Science 14, 2247–2249 (2014).
32. Zhao, C. et al. Changes of corneal endothelium in diabetes patients after cataract phacoemulsification surgery by confocal
microscopy. International Eye Science 13, 876–879 (2013).
33. Wang, B., Li, J. X., Wang, Y. L., Wu, B. G. & Huo, J. X. Clinical effect analysis of phacoemulsification on cataract patients with diabetes
mellitus. International Eye Science 13, 1163–1166 (2013).
34. Yang, R. et al. The influence of phacoemulsification on corneal endothelial cells at varying blood glucose levels. Eye science 26,
91–95, https://doi.org/10.3969/j.issn.1000-4432.2011.02.018 (2011).
35. Hugod, M. et al. Corneal endothelial cell changes associated with cataract surgery in patients with type 2 diabetes mellitus. Cornea
30, 749–753, https://doi.org/10.1097/ICO.0b013e31820142d9 (2011).
36. Wu, L. A., Zhang, L., Wang, C. Y. & Yang, X. G. Study on corneal endothelial cells after phacoemulsification in diabetic cataract.
International Journal of Ophthalmology 10, 1290–1293 (2010).
37. Wu, Z. D., Zhong, J. X., Mai, S. L., Zheng, M. & Zheng, K. Influence of cataract phacoemulsification on corneal endothelial cells in
diabetes. International Journal of Ophthalmology 8, 1908–1909 (2008).
38. Inoue, K. et al. Corneal endothelial cell morphology in patients undergoing cataract surgery. Cornea 21, 360–363 (2002).
39. Lee, J. S., Oum, B. S., Choi, H. Y., Lee, J. E. & Cho, B. M. Differences in corneal thickness and corneal endothelium related to duration
in diabetes. Eye (London, England) 20, 315–318, https://doi.org/10.1038/sj.eye.6701868 (2006).
40. Ohguro, N., Matsuda, M., Ohashi, Y. & Fukuda, M. Topical aldose reductase inhibitor for correcting corneal endothelial changes in
diabetic patients. The British journal of ophthalmology 79, 1074–1077 (1995).
41. Herse, P. R. Corneal hydration control in normal and alloxan-induced diabetic rabbits. Investigative ophthalmology & visual science
31, 2205–2213 (1990).
42. Schultz, R. O., Matsuda, M., Yee, R. W., Edelhauser, H. F. & Schultz, K. J. Corneal endothelial changes in type I and type II diabetes
mellitus. American journal of ophthalmology 98, 401–410 (1984).
43. Joyce, N. C. Proliferative capacity of corneal endothelial cells. Experimental eye research 95, 16–23, https://doi.org/10.1016/j.
exer.2011.08.014 (2012).
44. Furuse, N., Hayasaka, S., Yamamoto, Y. & Setogawa, T. Corneal endothelial changes after posterior chamber intraocular lens
implantation in patients with or without diabetes mellitus. The British journal of ophthalmology 74, 258–260 (1990).
45. Dhasmana, R., Singh, I. P. & Nagpal, R. C. Corneal changes in diabetic patients after manual small incision cataract surgery. Journal
of clinical and diagnostic research: JCDR 8, Vc03–vc06, https://doi.org/10.7860/jcdr/2014/7955.4288 (2014).
46. Schultz, R. O., Glasser, D. B., Matsuda, M., Yee, R. W. & Edelhauser, H. F. Response of the corneal endothelium to cataract surgery.
Archives of ophthalmology (Chicago, Ill.: 1960) 104, 1164–1169 (1986).
47. Mathew, P. T., David, S. & Thomas, N. Endothelial cell loss and central corneal thickness in patients with and without diabetes after
manual small incision cataract surgery. Cornea 30, 424–428, https://doi.org/10.1097/ICO.0b013e3181eadb4b (2011).
48. Wan, B. B. & Xu, J. Changes of corneal endothelial cell after phacoemulsification for patients with different preoperative level of
HbA1c. International Eye Science 15, 1158–1160 (2015).
49. Kotecha, A. et al. Corneal biomechanical characteristics in patients with diabetes mellitus. Journal of cataract and refractive surgery
36, 1822–1828, https://doi.org/10.1016/j.jcrs.2010.08.027 (2010).
50. Marano, C. W. & Matschinsky, F. M. Biochemical manifestations of diabetes mellitus in microscopic layers of the cornea and retina.
Diabetes/metabolism reviews 5, 1–15 (1989).
51. Narayanan, S. Aldose reductase and its inhibition in the control of diabetic complications. Annals of clinical and laboratory science
23, 148–158 (1993).
52. Jacot, J. L., Hosotani, H., Glover, J. P., Lois, N. & Robison, W. G. Jr. Diabetic-like corneal sensitivity loss in galactose-fed rats
ameliorated with aldose reductase inhibitors. Journal of ocular pharmacology and therapeutics: the official journal of the Association
for Ocular Pharmacology and Therapeutics 14, 169–180, https://doi.org/10.1089/jop.1998.14.169 (1998).
53. Murata, T. et al. The relationship between accumulation of advanced glycation end products and expression of vascular endothelial
growth factor in human diabetic retinas. Diabetologia 40, 764–769 (1997).
54. Takahashi, H. et al. Matrix metalloproteinase activity is enhanced during corneal wound repair in high glucose condition. Current
eye research 21, 608–615 (2000).
55. Woessner, J. F. Jr. Matrix metalloproteinases and their inhibitors in connective tissue remodeling. FASEB journal: official publication
of the Federation of American Societies for Experimental Biology 5, 2145–2154 (1991).
56. Do, J. R., Oh, J. H., Chuck, R. S. & Park, C. Y. Transient corneal edema is a predictive factor for pseudophakic cystoid macular edema
after uncomplicated cataract surgery. Korean journal of ophthalmology: KJO 29, 14–22, https://doi.org/10.3341/kjo.2015.29.1.14
(2015).
57. Chidambaram, P. Corneal Biomechanics as a Function of Race, The Ohio State University, (2017).
58. Leite, M. T. et al. Comparison of corneal biomechanical properties between healthy blacks and whites using the Ocular Response
Analyzer. American journal of ophthalmology 150, 163–168, e161 (2010).
59. Higgins, j. P. & Green, S. Cochrane Handbook for Systematic Reviews of Interventions (2011).

Acknowledgements
This study was supported by Program of National Natural Science Foundation of China (No. 81371001),
Program of National Natural Science Foundation (No. 81570822), Zhejiang Key LaboratoryFund of China (No.
2011E10006), Project of National Clinical Key Discipline of Chinese Ministry of Health, Zhejiang Province Key
Research and Development Program (2015C03042).

Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7 15

www.nature.com/scientificreports/

Author Contributions
K.Y. and Y.Z.T. designed this study. Y.Z.T., X.Y.C., X.B.Z., Q.M.T. and S.Y.L. collected and double checked the
data. Y.Z.T. and X.Y.C. carried out the statistical analysis. Y.Z.T. wrote the paper. K.Y. and X.Y.C. provided critical
revision to the article. All author participated in revision and approved the final version for submission.

Additional Information
Supplementary information accompanies this paper at https://doi.org/10.1038/s41598-017-14656-7.
Competing Interests: The authors declare that they have no competing interests.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Cre-
ative Commons license, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons license and your intended use is not per-
mitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

© The Author(s) 2017

Scientific REPOrTS | 7: 14128 | DOI:10.1038/s41598-017-14656-7 16